Optimization Research And Application Of DGA Algorithm For Power Transformer

With the rapid development of modern power grid and the widespread application of power transformer in power grid,as the core equipment of power grid,the operation quality of transformer is related to the operation reliability of the whole power grid,especially its repair is difficult and time consuming.In order to ensure the economy,stability and safety of power system operation,it is necessary to attach great importance to transformer fault diagnosis.Oil-immersed transformers are widely used in high-voltage and large-capacity power transformers at home and abroad,and dissolved gas analysis in oil(DGA)is an effective method for early detection and prevention of transformer faults.On the basis of extensive collection of transformer operation fault data,in order to overcome and improve the misjudgment when the ratio is just at the boundary of coding interval in the application of three-ratio diagnosis method.On the basis of BP network structure,principle,standard BP algorithm and improved BP algorithm,the BP neural network has been applied to transformer fault diagnosis,and achieved good classification results.In order to improve the BP neural network’s tendency to fall into local minima and slow convergence speed,a PSO-BP algorithm which organically combines BP algorithm and PSO algorithm is proposed,that is,the PSO algorithm is fully used to effectively optimize the threshold and weight of BP network,and the optimization is obtained.Experimental data analysis shows that PSO-BP network improves the accuracy of fault diagnosis.Based on the advantages of data-based statistical learning theory of Support Vector Machine(SVM),this paper applies SVM to diagnose transformer faults based on DGA.Firstly,a method of constructing transformer fault diagnosis model based on support vector machine is proposed,and the parameters of the model are determined by network search method.The PSO algorithm is used to optimize the parameters of SVM,the radial basis function(RBF)kernel function parameters and error items in the fault diagnosis model,and the accuracy of classifier is selected as the fitness function of PSO to improve the accuracy of SVM algorithm.Based on the results of case analysis,it can be seen that the diagnosis efficiency,classification processing ability and application promotion ability of the constructed transformer fault diagnosis model are very significant,which can provide a reliable and effective path for the related work of transformer fault diagnosis.